Structure-based macrocyclization yields hepatitis C virus NS5B inhibitors with improved binding affinities and pharmacokinetic properties

Article abstract of DOI:10.1002/anie.201200110

The concept of drug-likeness distills the physicochemical properties of small-molecule drugs to a set of rules. Macrocyclic drugs are known to break these rules. A structure-based macrocyclization strategy was applied to design new hepatitis C virus NS5B inhibitors with improved pharmacokinetic properties, exemplifying a rational strategy for overcoming the confines of standard drug-like chemical space . Copyright

Full text of DOI:10.1002/anie.201200110

Angewandte
Chemie
DOI: 10.1002/anie.201200110
Small-Molecule Inhibitors
Structure-Based Macrocyclization Yields Hepatitis C Virus NS5B
Inhibitors with Improved Binding Affinities and Pharmacokinetic
Properties
Maxwell D. Cummings,* Tse-I Lin, Lili Hu, Abdellah Tahri, David McGowan, Katie Amssoms,
Stefaan Last, Benoit Devogelaere, Marie-Claude Rouan, Leen Vijgen, Jan Martin Berke,
Pascale Dehertogh, Els Fransen, Erna Cleiren, Liesbet van der Helm, Gregory Fanning,
Kristof Van Emelen, Origꢀne Nyanguile, Kenny Simmen, Pierre Raboisson, and
Sandrine Vendeville*
Macrocyclization is one molecular design strategy that can be
used to improve the drug-related characteristics of small
molecules and peptides,[1–4] and thus may provide a route to
a less restrictive “drug-likeness space”. Structure-based
design is typically focused on the optimization of a chemical
series to increase affinity toward a target protein: modifica-
tion to gain additional binding contacts and rigidification to
favor a bio-active conformation are two common strategies.
In addition to affinity for the primary target, a 3D structure of
a relevant protein–small molecule complex can be used to
guide the optimization of other pharmaceutical properties.
Substituent positions that project away from the binding site
may be chemically exploited to alter one or more PK- or
ADME-related properties (e.g. “solubilizing groups”) while
having minimal effect on the therapeutically relevant binding
interaction (PK = pharmacokinetics; ADME = absorption,
distribution, metabolism, and excretion). Pursuing a design
strategy based on publicly available 3D structures, we
combined rigidification and the introduction of a PK-enhanc-
ing group in a single modification by forming a macrocycle
between two solvent-exposed substituent positions of a small-
molecule inhibitor of an essential viral enzyme. This approach
yielded new inhibitors with improved PK properties as well as
increased enzyme affinities and antiviral activities, while
appearing less drug-like according to standard calculated
criteria.
A series of indole-based inhibitors is known to bind to
a cryptic site of the thumb domain of the NS5B RNA
polymerase (NS5B) of hepatitis C virus (HCV) remote from
the active site,[5–7] and crystal structures of three early indole
inhibitors bound to this site are available from the protein
data bank (PDB) (Figure 1).[5,6] In these structures the
cyclohexyl ring and one edge of the phenyl and indole rings
project into the hydrophobic binding pocket, while the more
polar indole N-substituents and the carboxylate are largely
solvent-exposed. The exposed carboxylate forms a salt bridge
with Arg503 at the edge of the binding pocket, an important
binding interaction (Figure 1).[8]
Several groups have studied the indole series of NS5B
inhibitors, exploring bioisosteres of the parent carboxylate
(e.g. amide, acylsulfonamide, acylsulfamide) as well as
introducing solvent-exposed permeability- and/or PK-
enhancing substituents on the indole N.[6,8–10] Combined with
scaffold rigidification, these approaches have yielded new
tetracyclic (e.g. 2DXS in Figure 1) and pentacyclic systems
with enhanced binding affinities.[6,7,11] Unfortunately, the
resultant compounds have been burdened with pharmaceuti-
cally undesirable physico-chemical properties or one or more
undesirable structural features. For example, carboxylates
were shown to produce potentially toxic or immunogenic acyl
glucuronides through a phase-two metabolic pathway,[12–14]
and such acyl glucuronides have been observed with the early
leads 4[8,15] and 5[7] (Figure 2). Furthermore, the permeability/
PK-enhancer strategy evolved almost exclusively to zwitter-
[*] Dr. M. D. Cummings, Dr. T. Lin, L. Hu, Dr. A. Tahri, D. McGowan,
Dr. K. Amssoms, S. Last, Dr. B. Devogelaere, M.-C. Rouan,
Dr. L. Vijgen, Dr. J. M. Berke, P. Dehertogh, E. Fransen, E. Cleiren,
L. van der Helm, Dr. G. Fanning, Dr. K. Van Emelen,
Prof. Dr. O. Nyanguile, Dr. K. Simmen, Dr. P. Raboisson,
Dr. S. Vendeville
Janssen Infectious Diseases BVBA (formerly Tibotec BVBA)
Turnhoutseweg 30, 2340 Beerse (Belgium)
E-mail: svendev1@its.jnj.com
Dr. M. D. Cummings
Janssen Research & Development
Welsh & McKean Roads, Spring House, PA 19477 (USA)
E-mail: mcummin1@its.jnj.com
ionic molecules possessing
a
solvent-exposed basic
amine,[7,8,16] a feature that could cause high inter-patient
variability in absorption.[17]
Dr. B. Devogelaere
Biocartis NV, Mechelen (Belgium)
We hypothesized that we could address the PK limitations
of the indoles by macrocyclization of the solvent-exposed
inhibitor moieties. An initial set of macrocycles was designed
and evaluated by molecular modeling and conformational
analysis.[18] Overlay of the three publicly available complex
structures indicated that macrocyclization through a tether
between the carboxylate and either the indole N or the
bridged-phenyl moiety was consistent with the observed
Prof. Dr. O. Nyanguile
University of Applied Sciences, Institute of Life Technologies
Rte du Rawyl 47, 1950 Sion 2 (Switzerland)
Supporting information for this article (full references, protein
crystallography, and inhibitor synthesis and characterization) is
available on the WWW under http://dx.doi.org/10.1002/anie.
201200110.
Angew. Chem. Int. Ed. 2012, 51, 4637 –4640
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4637

.
Angewandte
Communications
Figure 2. Indole inhibitors 4[7,8] and 5.[7]
substituents connected by acylsulfamide-containing tethers in
a 17- or 16-membered macrocycle, respectively (Figure 3).
We determined the 2.1 ꢀ resolution crystal structure of 7
bound to NS5B.[20] Bound 7 occupies the expected thumb
domain site (Figure 3).[21] The observed binding mode of the
tetracyclic core of the designed macrocyclic inhibitor is fully
consistent with the three public structures of non-macrocyclic
inhibitor complexes that formed the basis of our structure-
based design strategy (Figure 1).[22] The macrocycle linker is
directed away from the protein surface, with the constrained
acylsulfamide replacing the carboxylate of previous inhib-
itors, forming two hydrogen bonds with the guanidine of
Arg503, mimicking the contacts observed in previous NS5B-
indole complexes (Figure 3).[5,6]
Previous structural and mechanistic studies indicate a key
function for the fingertip L1 loop–thumb domain interac-
tion.[5,23] Binding of indole inhibitors blocks the fingertip–
thumb interaction, inhibiting polymerase activity by, appa-
rently, locking NS5B in an inactive “open” conformation.[5,23]
In past complexes,[5,6] residues 22–35 of the displaced loop
have been disordered. Interestingly, the 7–NS5B complex
reveals new structural information about the binding site,
with four of the loop residues (residues 22 and 33–35)
disordered in previous structures now resolved; in the new
structure residues 33–35 form a helical turn positioned to
more fully enclose the methoxyphenyl ring of bound 7
(Figure 3b). The more enclosed binding site and presumably
smaller protein conformational change observed for the 7–
NS5B complex may be due to a specific (but weak, given the
observed distance) intermolecular hydrogen bond between
the macrocycle carbonyl of 7 and His34 (Figure 3c). Such
a contact also seems likely for 6, but is unlikely or impossible
for most other reported indole and benzimidazole NS5B
inhibitors, assuming a similar binding mode.
Additional binding interactions and/or different confor-
mational adjustments (protein and/or inhibitor) required for
binding may explain the higher binding affinities and lower
off-rates of macrocycles 6 and 7, compared to the acyclic
predecessor 4 (Table 1). The consequent longer inhibition
half-lives of the macrocycles may be relevant to therapeutic
efficacy.[24] Enzymatic and replicon potencies do not fully
parallel binding affinities; high protein binding (> 99% in
human serum for both macrocycles) rather than limited
permeability may account for this discrepancy, and explains
the 8- to 9-fold shift in EC50 measured under high serum
conditions (Table 1). Rather than trying to minimize the
Figure 1. Indole inhibitors of HCV NS5B. a) Overlay of the three early
inhibitors with structures available from the PDB (2BRK green color-
by-atom, 2BRL cyan color-by-atom, 2DXS magenta color-by-atom),
based on superposition of conserved secondary structure elements in
the region of the indole binding site. b) Inhibitor chemical structures,
PDB ID codes of the complex structures and various properties of the
small-molecule inhibitors (MW=molecular weight in gmolÀ1
;
clogP=logarithm of the partition coefficient between n-octanol and
water; HD=number of hydrogen-bond donors; HA=number of
hydrogen-bond acceptors). c) General macrocyclization design
schemes.
indole binding mode (Figure 1). Attractively, this approach
also maintained the contact with Arg503 while restraining the
solvent-exposed inhibitor moieties. Our macrocyclization
strategy was initially validated by comparing open and
cyclized amide analogues.[19] Likely carboxylate replace-
ments[10] were tested in our macrocyclization strategy to
optimize the key polar interaction with Arg503.[19] Previous
structural studies investigating the biaryl torsion[6] and our
own explorations[18] led us to focus on the tetracyclic indole-
based system. Subsequent lead optimization focused on the
more potent acylsulfamide derivatives combined with the
more rigid tetracyclic macrocycle template, exploring macro-
cycle size, substituents on nitrogens and on the phenyl ring,
and the linker between the phenyl and the indole rings.[18] Our
design strategy quickly showed promise, yielding molecules
with enhanced binding affinities and improved pharmaco-
kinetic profiles,[18] and led to the discovery of inhibitors 6 and
7 (see Supporting Information for inhibitor synthesis and
characterization), based on a tetracyclic core with two amide
4638 www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4637 –4640

Angewandte
Chemie
Table 1: In vitro profile of macrocyclic indoles and the acyclic carboxylic
acid 4.
Serum shift[d]
(-fold)
[b]
[c]
Cmpd.
Kd
koff
IC50
EC50
[nm][a]
[sÀ1 [a]
]
[nm]
[nm]
4
6
7
136
4.0
2.4
380ꢀ10À5
10.0ꢀ10À5
4.8ꢀ10À5
48[e]
31
26
260
81
75
8
9
8
[a] Dissociation constant (Kd) and off-rate constant (koff) determined by
surface plasmon resonance with NS5B con1b(D21). [b] Inhibition
potency on the NS5BD21 polymerase (20 nm). [c] Replicon potency.
[d] Ratio of the replicon potency (EC50) measured in the presence of 40%
human serum divided by that observed in the absence of human serum.
[e] From ref. [25].
serum shift (only 3-fold decrease in EC50) and good bioavail-
ability across different species (49% in rat, 78% in dog), but
shows limited liver distribution in rat (4 ꢁ EC50 at 50% human
serum).[25] Because it is anticipated that liver distribution and
liver/plasma ratio at 7 h post-dose are important contributors
to therapeutic efficacy,[26,27] 6 was selected for clinical devel-
opment.
Structure-based macrocyclization has led to new inhib-
itors with improved pre-clinical profiles. We see four key
distinctions between the present macrocycles and their non-
macrocyclic predecessors. First, the carboxylate of the non-
macrocyclic analogues has been replaced by a cyclized
acylsulfamide, maintaining specific contact with the guanidine
of Arg503 (Figure 3) while eliminating the potential for
glucuronidation, a liability for carboxylate-containing drugs.
Second, our strategy overcame the requirement for the
presence of a basic amine—essential for cellular replicon
potency in non-macrocyclic series, and avoids zwitterionic
structures—a pharmaceutically undesirable feature.[7,25]
Third, binding of 7 recruits new elements of the inhibitor
binding site, providing additional information useful for
structure-based design of inhibitors targeted to this site.
Fourth, and perhaps most interesting, the macrocyclic inhib-
itors show improved PK behavior while appearing less “drug-
like” than their non-macrocyclic predecessors, according to
standard calculated physicochemical criteria of “drug-like-
ness” (Figure 1–3). Recent analyses indicate that current drug
discovery efforts are often focused on molecules at or beyond
the frontier of what is generally described as “drug-like”
chemical space.[28–30] This is not, however, entirely unreason-
able, since many useful drugs exceed the limits typically
implied by the term “drug-like”.[28,29] Chemistry strategies
that facilitate exploration of the limits of what is acceptable
chemical space for drugs should lead to improved under-
standing of the “drug-like” concept, and should allow drug
discovery chemists to work with a broader molecular palette.
One approach adopted by nature for achieving larger small-
molecule bio-actives is the macrocycle;[1,4] the value of this
approach is underscored by the prevalence of macrocycles in
bio-active natural products and natural product-derived
drugs. Various factors are thought to contribute to the
improved drugability of macrocyclic small molecules, includ-
ing restricted flexibility, reduced metabolism (esp. for cyclized
peptides), and shielding of polar atoms through intramolec-
ular hydrogen bonding.[1,3,4] Rather than being property
Figure 3. Macrocyclic indole inhibitors. a) Chemical structures of 6 and
7. b) 2.1 ꢁ resolution crystal structure of 7 (orange color-by-atom;
structure deposited as PDB ID 4DRU) bound to genotype 1b HCV
NS5B polymerase, with early indole inhibitors overlaid as in Figure 1a
and orange surface indicating the additional residues resolved in the
new structure (see text). c) Interaction with Arg503 and His34 for 7
and the indole inhibitor from PDB ID 2DXS (overlay as in Figure 1a;
distances shown in ꢁ).
observed serum shift, we focused on optimization of the
[liver]/[serum shifted potency] ratio to maximize in vivo
efficacy.
As well as affecting binding affinities and antiviral
potencies, macrocyclization yielded inhibitors with improved
PK profiles. These compounds show high bioavailability, good
liver distribution and moderate clearance, with liver concen-
trations 7 h post-dose more than 10 ꢁ greater than the
measured serum-shifted potencies in rat (Table 2). Con-
versely, after extensive optimization efforts most non-macro-
cyclic compounds were limited by poor bioavailability (e.g. 4:
F = 10%), resulting in poor liver distribution (e.g. 4: 1.1 ꢁ
EC50 at 50% human serum 6 h post-dose (at 3 mgkgÀ1)).[8]
Although bioavailability could be improved, more optimized
non-macrocyclic inhibitors still suffer from poor liver distri-
bution. For example, MK-3281, a compound that entered
clinical development, is potent in the replicon with limited
Angew. Chem. Int. Ed. 2012, 51, 4637 –4640
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 4639

.
Angewandte
Communications
Table 2: Pharmacokinetic parameters of macrocyclic analogues 6 and 7 in male Sprague–Dawley rats (n=3) and macrocycle 6 in Beagle dogs
(n=3).[a]
Cl[b]
tmax
[h]
Cmax
AUCinf
[mmh]
Liver[d]
[mm]
L/P[e]
L/EC50
F[g]
[%]
[c]
[c]
[c]
Species
Cmpd.
[LhÀ1 kgÀ1
]
[mm]
(40% HS)[f]
rat
rat
dog
6
7
6
3.2
2.4
0.54
2.7
0.8
1.2
0.7
1.3
10.2
4.8
7.0
25.8
13.0
8.3
–
46
23
–
18
12
–
>66
76
87
[a] 2 mgkgÀ1 intravenous, 10 mgkgÀ1 peroral; intravenous vehicle PEG400/saline 70/30; peroral vehicle PEG400/2% vitE-TPGS. [b] Plasma clearance.
[c] tmax =time to maximum plasma concentration, Cmax =maximum plasma concentration, AUCinf =area under the curve extrapolated to infinite time.
[d] Liver concentrations after oral administration 7 h post-dosing. [e] Liver/plasma ratio after oral administration 7 h post-dosing. [f] Liver/EC50 (at
40% human serum) ratio. [g] Oral bioavailability.
[17] Y. H. Zhao, M. H. Abraham, J. Le, A. Hersey, C. N. Luscombe,
outliers, these more complex small-molecules may simply be
G. Beck, B. Sherborne, I. Cooper, Pharm. Res. 2002, 19, 1446.
[18] Extensive structure–activity details will be published elsewhere.
revealing the limitations of standard 2D property calculations.
While not unheard of, macrocycles are considered under-
[19]
exploited in current synthetic drugs;[1] however, recent
advances in relevant synthetic chemistry techniques, most
notably the pioneering work of Grubbs,[31,32] have opened the
door to broader exploration of synthetic small-molecule
macrocycles. In the present work we have used structure-
based design to guide macrocyclization, leading to the
discovery of new inhibitors that maintain potency while
showing substantially improved pharmacokinetic behavior.
Received: January 5, 2012
Revised: February 7, 2012
Published online: March 30, 2012
Keywords: hepatitis C virus · inhibitors · macrocycles ·
.
polymerases · pharmacokinetics
[1] E. M. Driggers, S. P. Hale, J. Lee, N. K. Terrett, Nat. Rev. Drug
Discovery 2008, 7, 608.
[20] DC21 form of HCV genotype 1b strain J4 NS5B.
[21] The inhibitor is bound to one of the two NS5B monomers in the
crystallographic asymmetric unit; in the other monomer the
fingertip loop displacement required for inhibitor binding is
blocked by a crystallographic contact.
[22] The cyclohexyl ring of the new inhibitor is the most deeply
buried part of the bound inhibitor, and the bridged indole-
phenyl torsion adopts the dihedral angle observed previously in
both free and rigidified systems.[6]
[23] B. K. Biswal, M. M. Cherney, M. Wang, L. Chan, C. G. Yanno-
poulos, D. Bilimoria, O. Nicolas, J. Bedard, M. N. James, J. Biol.
Chem. 2005, 280, 18202.
[24] 110 min (6) and 242 min (7), compared to 3 min for the 4–NS5B
complex. In some therapeutic contexts increased inhibitor
residence time may be of pharmacological benefit.[33]
[25] F. Narjes et al., J. Med. Chem. 2011, 54, 289.
[26] D. Lamarre et al., Nature 2003, 426, 186.
[27] N. J. Liverton et al., J. Am. Chem. Soc. 2008, 130, 4607.
[28] A. Alex, D. S. Millan, M. Perez, F. Wakenhut, G. A. Whitlock,
Med. Chem. Commun. 2011, 2, 669.
[29] P. D. Leeson, B. Springthorpe, Nat. Rev. Drug Discovery 2007, 6,
881.
[30] W. P. Walters, J. Green, J. R. Weiss, M. A. Murcko, J. Med. Chem.
2011, 54, 6405.
[31] G. C. Vougioukalakis, R. H. Grubbs, Chem. Rev. 2010, 110, 1746.
[32] L. Yet, Chem. Rev. 2000, 100, 2963.
[2] E. Marsault, M. L. Peterson, J. Med. Chem. 1961, 54, 1961.
[3] T. Rezai, J. E. Bock, M. V. Zhou, C. Kalyanaraman, R. S. Lokey,
M. P. Jacobson, J. Am. Chem. Soc. 2006, 128, 14073.
[4] L. A. Wessjohann, E. Ruijter, D. Garcia-Rivera, W. Brandt, Mol.
Diversity 2005, 9, 171.
[5] S. Di Marco et al., J. Biol. Chem. 2005, 280, 29765.
[6] K. Ikegashira et al., J. Med. Chem. 2006, 49, 6950.
[7] I. Stansfield, C. Ercolani, A. Mackay, I. Conte, M. Pompei, U.
Koch, N. Gennari, C. Giuliano, M. Rowley, F. Narjes, Bioorg.
Med. Chem. Lett. 2009, 19, 627.
[8] S. Harper, J. Med. Chem. 2005, 48, 4547.
[9] P. L. Beaulieu, J. Gillard, D. Bykowski, C. Brochu, N. Dansereau,
J.-S. Duceppe, B. Hachꢂ, A. Jakalian, L. Lagacꢂ, S. LaPlante, G.
McKercher, E. Moreau, S. Perreault, T. Stammers, L. Thauvette,
J. Warrington, G. Kukolj, Bioorg. Med. Chem. Lett. 2006, 16,
4987.
[10] I. Stansfield, M. Pompei, I. Conte, C. Ercolani, G. Migliaccio, M.
Jairaj, C. Giuliano, M. Rowley, F. Narjes, Bioorg. Med. Chem.
Lett. 2007, 17, 5143.
[11] J. Habermann, E. Capito, M. del R. R. Ferreira, U. Koch, F.
Narjes, Bioorg. Med. Chem. Lett. 2009, 19, 633.
[12] M. J. Bailey, R. G. Dickinson, Chem.-Biol. Interact. 2003, 145,
117.
[13] M. Castillo, P. C. Smith, Drug Metab. Dispos. 1995, 23, 566.
[14] C. Skonberg, J. Olsen, K. G. Madsen, S. H. Hansen, M. P. Grillo,
Expert Opin. Drug Metab. Toxicol. 2008, 4, 425.
[15] C. Giuliano et al., Xenobiotica 2005, 35, 1035.
[16] P. L. Beaulieu et al., Bioorg. Med. Chem. Lett. 2010, 20, 857.
[33] R. A. Copeland, D. L. Pompliano, T. D. Meek, Nat. Rev. Drug
Discovery 2006, 5, 730.
4640 www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 4637 –4640